> Rui Sousa <[EMAIL PROTECTED]> wrote:
> Roland Orre wrote:
>> It seems as my "stuck on TLB IPI wait [CPU#1]" problems disappeared
>> when I changed from OSS to ALSA.
>>
>> After upgrading the system to Debian potato (X and Gnome are the
>> latest from unstable) and installing the kernel 2.2.16 (Debian call it
>> 2.2.17) from source I got a lot of problem with spontaneous crashing
>> and also the message "kernel: stuck on TLB IPI wait [CPU#1]" now and
>> then, when the machine was not responding for some seconds.
> Rui:
> Were you using the SB Live driver at this point? Which version (from the
> kernel source, the CVS version at creative.opensource.com, ...)
Roland:
I used the commercial OSS (osslinux393n-240-test5-SMP.tar.gz) driver
for SB Live. I was then so amazed and satisfied by the multi open
property of SB Live so I installed a SB Live on my other systems too.
I was, however, somewhat dissapointed that it didn't manage the
synthesizer and then I read some more about ALSA and found that
it seemed as they actually do manage the synth.
>> To confirm that it was a problem with my system I installed an identical
>> system on another machine (almost same, one is using ASUS P2BS and the
>> other one uses ASUS P2LS, both with 512 Mb ECC and scsi-disks), I got
>> the same problem on that machine.
>> I then upgraded one machine to 2.4.0-test5 when I got rid of
>> the "kernel: stuck on TLB IPI wait [CPU#1]"-message, but...
>> I still had problems with spontaneous crashing.
> Rui:
> There is a known deadlock in this driver. I'll submit a patch with
> just this fix since the big patch I've submitted a couple of times
> before (also fixing this problem) never got applied...
Roland:
Oh, great! I look forward to that.
At the moment I am running the two (now 2.4.0-test6 kernels) in
parallell, one with the OSS emu101k driver included in the kernel and
one with the ALSA-driver. Now I've also installed a serial console,
enabled the CONFIG_MAGIC_SYSRQ (to avoid tedious fsck's in case of
new freezes) the nmi-watchdog and also installed the the microcode
upgrade.
One of the machines is a backup for a combined file/mail/news-server
and workstation (still running on 2.0.36 and no IO_APIC, but with
and uptime around 240 days). It will probably take some time before
I consider the new system stable enough to replace the 2.0.36...
The last night, after the machine (running ALSA) had frozen the
2.4.0-test5 system and I had rebooted when I started gdm I could
not login, which was a fairly interresting error. The machine
didn't freeze as usual (about once a day lately... (both))
This machine has become somewhat stable since I turned the
gnome sound events off. With somewhat stable I mean that it had
been running since morning with no freezing. Each freezing I have
had has been associated with a mouse event as far as I can understand.
I've also suspected the tagged queing so I've turned this off.
Now the machine freezed again and after reboot and start of gdm
I didn't get any response to my keyboard.
This time I could login to the machine from another machine and
just for testing I played some sound. Then the text I had written
in the gdm login window was echoed and the mouse pointer moved.
I then checked the interrupt table and noticed that the mouse
this time had been put on the same interrupt as the EMU10K1.
The mouse use to be edge triggered but now it had been considered
level triggered together with the EMU10K1 board.
laplace:/var/log# cat /proc/interrupts
CPU0 CPU1
0: 221788 235005 IO-APIC-edge timer
1: 610 373 IO-APIC-edge keyboard
2: 0 0 XT-PIC cascade
9: 81066 81051 IO-APIC-level acpi, aic7xxx
10: 2579 2862 IO-APIC-level eth0
12: 440 423 IO-APIC-level EMU10K1, PS/2 Mouse
13: 1 0 XT-PIC fpu
NMI: 456717 456722
LOC: 456719 456717
ERR: 0
The board in the above case is ASUS PL97-DS.
After forcing the soundcard slot to interrupt 5 and forcing the
mouse-interrupt 12 to an ISA interrupt and telling that the OS
is a non-PNP-os (which I hadn't done on this machine) I got this
layout of the interrupts and the machine worked again.
CPU0 CPU1
0: 1080401 1019310 IO-APIC-edge timer
1: 2436 2427 IO-APIC-edge keyboard
2: 0 0 XT-PIC cascade
5: 86948 85786 IO-APIC-level EMU10K1
9: 48209 49067 IO-APIC-level aic7xxx
10: 159400 162937 IO-APIC-level eth0
12: 15703 14190 IO-APIC-edge PS/2 Mouse
13: 1 0 XT-PIC fpu
NMI: 2099640 2099633
LOC: 2099658 2099657
ERR: 0
Best regards
Rolad Orre
-
Linux SMP list: FIRST see FAQ at http://www.irisa.fr/prive/dmentre/smp-howto/
To Unsubscribe: send "unsubscribe linux-smp" to [EMAIL PROTECTED]
